Oceanic Calcium Changes from Enhanced Weathering During the Paleocene‐Eocene Thermal Maximum: No Effect on Calcium‐Based Proxies N

PALEOCEANOGRAPHY, VOL. 26, PA3211, doi:10.1029/2010PA001979, 2011

Oceanic calcium changes from enhanced weathering during the Paleocene‐Eocene thermal maximum: No effect on calcium‐based proxies N. Komar1 and R. E. Zeebe1 Received 15 April 2010; revised 22 April 2011; accepted 11 May 2011; published 10 August 2011.

[1] During the Paleocene‐Eocene thermal maximum (PETM ∼55 Myr ago), prominent climatic and biogeochemical changes took place in the atmosphere, ocean, and on land. For example, deep‐sea temperatures rose by 5°C to 6°C, while sea surface temperatures at high latitudes increased by up to 9°C. In the sedimentary record, the onset of the PETM is marked by widespread dissolution of calcium carbonate on the seafloor. In addition, there is evidence for globally higher humidity, precipitation and increased weathering during the PETM. Both calcium carbonate dissolution and enhanced weathering probably affected the seawater calcium concentration. Here we investigate implications that possible changes in the ocean’s calcium inventory may have had on boron/calcium (B/Ca) and magnesium/calcium (Mg/Ca) ratios, which are used as proxies for deep water carbonate chemistry and temperature, respectively. We also examine effects on d44Ca of seawater, which is used as an indicator for variations in the marine calcium cycle. We focus on the magnitude of change in the ocean’s calcium ion concentration as a result of the carbon perturbation, which resulted in increased weathering fluxes and the dissolution of calcite on the ocean floor during the PETM. Different ranges of carbon input scenarios and their effect on ocean chemistry were examined using the Long‐term Ocean‐atmosphere‐Sediment CArbon cycle Reservoir (LOSCAR) model. We found that under the most plausible scenario, the calcium ion concentration change (D[Ca2+]) was less than 0.7% and around 2% in the most extreme scenario. Our results show that B/Ca and Mg/Ca proxies were not affected within analytical precision by changes in oceanic calcium due to weathering and carbonate dissolution during the PETM. The most extreme scenario (D[Ca2+] = 2%) would result in ∼4 mmol kg−1 uncertainty in 2− reconstruction of D[CO3 ]. The same scenario affects the temperature reconstruction by ∼0.2°C. The effect on the ocean’s calcium isotope budget was insignificant as well, 44 resulting in Dd Casw of less than 0.05‰. Citation: Komar, N., and R. E. Zeebe (2011), Oceanic calcium changes from enhanced weathering during the Paleocene‐Eocene thermal maximum: No effect on calcium‐based proxies, Paleoceanography, 26, PA3211, doi:10.1029/2010PA001979.

1. Introduction mass balance equation it is possible to estimate the mass of carbon needed for a −2.5‰ carbon isotope excursion, given [2] The Paleocene‐Eocene thermal maximum (PETM) the isotopic composition of the source. In the past, three took place around 55 million years ago. During the event, different sources have usually been considered: carbon from sea surface temperatures increased by 5–9°C in low and 13 13 the mantle (d C=−5‰), carbon in organic matter (d C= high latitudes [e.g., Zachos et al., 2003; Sluijs et al., 2006]. − ‰ d13 − ‰ ‰ 25 ), and carbon as methane ( C= 60 ). Conversely, Kennett and Stott [1991] observed a 2.5 decline in the if the carbon input mass was known, it would be possible to d13C of calcium carbonate in foraminifera from the Southern constrain the sources because of their different carbon iso- Ocean. This decrease was most likely due to the introduc- tope compositions. The different sources have led to spec- tion of isotopically light carbon into the ocean‐atmosphere ulations concerning the mechanism. Some, such as volcanic system [e.g., Koch et al., 1992; Dickens et al., 1995; Zachos intrusion, imply that the carbon drives the warming. Others, et al., 2001; Dunkley Jones et al., 2010]. By using a simple such as destabilization of oceanic methane hydrates imply that the carbon release is a feedback that can exacerbate warming [Dickens et al., 1995; Dickens, 2000; Pagani et al., 1Department of Oceanography, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, Hawaii, USA. 2006]. Modeling of the carbon cycle during the PETM indicates that the initial carbon pulse was around 3,000 Pg C Copyright 2011 by the American Geophysical Union. [Zeebe et al., 2009]. The intrusion of such a large amount of 0883‐8305/11/2010PA001979

PA3211 1of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211 carbon into the ocean‐atmosphere system had a series of deciphering paleoenvironmental changes. Past changes in the 2− consequences that affected the chemistry and biology of the concentration of CO3 are not easy to quantify even though oceans [e.g., Zachos et al., 2005; Thomas, 2007; Webb many studies have been carried out to investigate it. The et al., 2009]. There are several lines of evidence indicat- difficulties are mostly due to the limitations of the different ing intensified weathering rates during the PETM. For methods (proxies based on carbonate dissolution, theory and instance, the osmium isotopic composition of seawater empirical relationships) [e.g., Yu and Elderfield, 2007]. became more radiogenic during this period, indicating an Nonetheless, Yu et al. [2007] have developed a new method to accelerated hydrologic cycle and consequently enhanced reconstruct deep water chemistry, which is based on boron/ weathering [Ravizza et al., 2001]. Furthermore, increased calcium (B/Ca) ratios in benthic foraminifera. They found a 2− kaolinite abundances during the recovery stages of the correlation between benthic foraminiferal B/Ca and D[CO3 ] PETM also indicate increased runoff rates [Kelly et al., (the degree of carbonate saturation), which provides a 2− 2005; Ravizza et al., 2001; Robert and Kennett, 1994]. paleoproxy for deep water [CO3 ]. Furthermore, Elderfield The increase in both temperature and pCO2 most likely led and Ganssen [2000] and many others [e.g., Rosenthal et al., to enhanced silicate and carbonate weathering fluxes, which 1997; Lea et al., 2000; Dekens et al., 2002; Zachos et al., increased the calcium input to the ocean: 2003] have applied another correlation in planktonic and benthic foraminifera, which also involves the calcium con- CaCO þ CO þ H O ! Ca2þ þ 2HCO ð1Þ 3 2 2 3 centration. The magnesium/calcium (Mg/Ca) ratio in foraminifera shows a temperature dependence and is thus being CaSiCO þ 2CO þ H O ! Ca2þ þ SiO þ 2HCO ð2Þ used widely as a proxy for determining past records of tem- 3 2 2 2 3 perature. In this study, we also examine the effect of the marine calcium concentration and increased weathering on [3] Another consequence of the large carbon input into the Ca isotope composition of seawater (d44Ca ), which is ‐ sw the ocean atmosphere system was extensive calcite disso- used as an indicator for variations in the marine calcium cycle ’ lution that took place on the ocean s floor. The dissolution [e.g., Fantle and DePaolo, 2005; Heuser et al., 2005; Griffith was due to the depletion of carbonate ion in seawater: et al., 2008; Fantle, 2010]. 2+ CO þ H O ! HCO þ Hþ ð3Þ [6] In the studies mentioned above, it is assumed that [Ca ] 2 2 3 in seawater remained constant over relatively short timescales because of the long residence time of calcium in the þ 2 H þ CO3 ! HCO3 ð4Þ ocean (1 to 2 Ma). In this study, we modeled the changes in the deep ocean chemistry during the PETM, primarily 2+ In general, the rate of deposition/dissolution of calcite to/from focusing on possible changes in the inventory of Ca which, sediments depends on the saturation state of seawater. The if large enough, could have affected the validity of the above 2+ 2− stability is related to the concentrations of Ca and CO3 via mentioned proxies. In order to determine whether enhanced the saturation state of the solution (W = [Ca2+] × [CO2−]/K* ), weathering and deep sea dissolution can significantly alter 3 sp 2+ where K*sp is the solubility product of the calcium carbonate the Ca inventory of the sea, we conducted simulations mineral in question (e.g., calcite or aragonite). Thus, depo- using a long‐term carbon cycle model [Zeebe et al., 2008; 2+ 2− sition/dissolution depends on [Ca ] and [CO3 ] in seawater Zachos et al., 2008; Zeebe et al., 2009]. We consider dif- and K*sp, which increases with depth. The equilibrium calcite ferent scenarios by exposing the system to a wide range of saturation horizon is the depth at which W = 1. Below this carbon inputs (1,000 Pg C to 6,000 Pg C) and weathering point dissolution proceeds and W < 1. With increasing depth, parameterizations, i.e., we effectively vary the magnitude of W decreases and the rate of dissolution increases. The depth at carbonate dissolution and weathering fluxes (see below). which dissolution is balanced by the rain of calcite to the sediments is known as the calcite compensation depth (CCD) 2. Dissolution: Back‐of‐the‐Envelope Calculation [e.g., Ridgwell and Zeebe, 2005]. [7] As previously mentioned, there are at least two [4] The intrusion of CO into the seawater during the 2 mechanisms (terrestrial weathering and dissolution from the PETM resulted in rapid shoaling of the CCD (in less than seafloor) that could have altered the ocean’s [Ca2+] during 10 ky) by more than 2 km in the Atlantic Ocean [Zachos 2+ − the PETM. Estimating the effect on [Ca ] that is caused et al., 2005]. The decrease in [CO2 ] resulted in increased 3 solely by dissolution is relatively easy and is explained in area of undersaturation producing a shallower CCD, insti- this section and Appendix A. If the initial position of the gating the dissolution of calcite and thus additional input of − CCD was at 3.5 km and if we assume that the entire ocean calcium to the ocean. The decline in CO2 (in combination 3 floor above that depth (shallower than 3.5 km) underwent with O depletion) affects the stability and production of 2 dissolution, we calculate that [Ca2+] would have changed by CaCO minerals, which may have affected benthic forami- 3 approximately 0.4% (assuming that porosity varies with % nifera during the PETM, causing major extinctions [Kennett CaCO ; see Appendix A). Thus, an estimate of the disso- and Stott, 1991; Thomas, 2007; Ridgwell and Schmidt, 2010]. 3 lution effect on the ocean’s calcium budget can be obtained [5] The increased weathering as well as dissolution of without the use of numerical models. However, the effects calcite from the seafloor might have caused significant of weathering on [Ca2+] are more complex and not easily changes in the ocean’s calcium inventory during the PETM. predictable a priori. It turned out that the transient changes In this study, we assess potential artifacts induced in Ca‐based in seawater [Ca2+] are controlled by a time‐dependent bal- proxies due to a changing ocean calcium inventory. There are ance between river input and sediment burial, where the several methods which use Ca‐based proxies as a tool for

2of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 1. Schematic view of the LOSCAR model. The arrows indicate SO deep water formation; A, I, P, and T stand for the Atlantic, Indian, Pacific, and Tethys ocean; H is a high latitude box; L, M, and D correspond to low, intermediate, and deep ocean boxes; “g” is gas exchange, and “w” is weathering. magnitude of change differs in response to carbonate and state from total CO2 and total alkalinity use algorithms as silicate weathering (see Discussion). Such features are dif- described by Zeebe and Wolf‐Gladrow [2001]. Routines that ficult to quantify based on a back‐of‐the‐envelope calcula- allow for variations in the Ca2+ and Mg2+ concentration of tion. Here we employ a numerical modeling approach to seawater, which most likely differed from modern values addressed both, effects of dissolution and weathering on the during the Paleocene‐Eocene are described by Tyrrell and seawater calcium concentration during the PETM. Zeebe [2004]. In our PETM simulations we used [Ca2+]= 20 mmol kg−1 and [Mg2+] = 30 mmol kg−1 as inferred from 3. Model Description fluid inclusions in marine halites [Lowenstein et al., 2001; Horita et al., 2002], rather than the modern values of [Ca2+]= [8] The model used in this study is the LOSCAR (Long‐ −1 2+ −1 ‐ ‐ 10 mmol kg and [Mg ] = 53 mmol kg [Tyrrell and term Ocean atmosphere Sediment CArbon cycle Reservoir) Zeebe, 2004]. (Note that our conclusions also hold at modern model (Figure 1). It is a modified version of the carbon seawater Ca and Mg concentrations, as additional simulations cycle model developed by Walker and Kasting [1992], showed). Warmer surface and bottom water temperatures in coupled to a sediment module [Zeebe and Zachos, 2007]. It the late Paleocene and Eocene also impact equilibrium and represents the ocean basin bathymetry of earlier geologic ‐ solubility constants. For example, the calcite saturation times, such as the Paleocene Eocene era. In addition to the concentration at a bottom water temperature of 14–17°C Atlantic (A), Indic (I) and Pacific (P) basin, the Tethys (T) during the PETM is quite different from the modern tem- ocean is also included. Each of these four ocean basins are ∼ ‐ perature of 2°C. In the model, we used bottom water tem- subdivided into three boxes, which correspond to the low peratures of 12°C and 16°C prior to and during the PETM, latitude surface ocean (L), the intermediate water masses respectively, and included the effects of changes in temper- (M) and the deep ocean (D). There is also an additional box ature, and [Ca2+] and [Mg2+] on the stoichiometric equilib- that represents the high latitude ocean (H). The model uses rium constants. realistic volumes of ocean basins based on Paleocene‐ [10] The model runs were performed under a suite of car- Eocene topography [Bice and Marotzke, 2002]. Prior to the bon emission scenarios, with an initial pulse ranging from PETM, deepwater formation in the model was prescribed to 1,000 to 6,000 Pg C (covering the range of scenarios dis- occur in the Southern Ocean (SO), consistent with obser- cussed in the literature [Dickens et al., 1995; Panchuk et al., vations [Thomas et al., 2003]. After the PETM main phase, 2008]) and a release time over 6 ky [Zeebe et al., 2009]. The a steady contribution of North Pacific Deep Water formation model was run for 200 ky in total, the time span during which was assumed (the SO source remains active but is reduced calcium changes are important (the onset of calcium change is relative to its pre‐event strength). After ∼70 ky, the NP source ‐ within this time range as well as the essential stage of the is shut off and SO formation returns to pre event levels. The carbonate ion recovery phase). Additionally, effects of sili- model includes biogeochemical cycles of total carbon, total cate weathering are observed on this timescale. Observations alkalinity, phosphate, oxygen, stable carbon isotopes, and indicate that atmospheric CO2, temperature and precipitation calcium. increased during the PETM as a result of the carbon pertur- [9] Ocean carbonate chemistry routines to calculate 2− bation, which enhances terrestrial weathering of carbonate parameters such as [CO2], [CO3 ], pH, and calcite saturation and silicate rocks [Berner et al., 1983; White and Blum, 1995;

3of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Ravizza et al., 2001; Schmitz and Pujalte, 2007]. In the over the first 6 ky, which is the most likely scenario according model, the carbonate (FC) and silicate (FSi) weathering fluxes to Zeebe et al. [2009]. Figure 2a shows our PETM carbon are parameterized as a function of atmospheric CO2. In the release scenario, where t = 0 corresponds to the onset of the model, this is expressed as: PETM (Paleocene‐Eocene boundary). It is important to note ÀÁ that besides this large initial pulse, an additional smaller pulse F ¼ F0 pCO =pCO0 nSi ð5Þ Si Si 2 2 was simulated, followed by a continuous carbon release (∼1,500 Pg C). This prolonged carbon release is necessary to ÀÁn simulate the observed duration of deep‐sea carbonate disso- F ¼ F0 pCO =pCO0 CC ð6Þ C C 2 2 lution [see Zeebe et al., 2009]. Without the additional release, 0 0 the model was unable to reproduce the carbon isotope where (FC) and (FSi) are initial carbonate and silicate excursion duration because d13C values returned to pre‐ weathering fluxes for the PETM, set at 15.8 × 1012 mol C yr−1 12 −1 excursion values too quickly [Zeebe et al., 2009]. Further- and 5.7 × 10 mol C yr , respectively (volcanic input equals more, the results of this simulation (a pulsed carbon release silicate weathering flux). The modern weathering carbonate rather than a single input peak) are more consistent with d13C flux is ∼12 × 1012 [Morse and Mackenzie, 1990] and silicate 12 records [Dickens, 2001; Bowen et al., 2004]. The prolonged and mantle/volcanic CO2 outgassing fluxes are ∼5×10 mol −1 carbon release is also important to simulate the observed Cyr [Walker and Kasting, 1992]. duration of deep‐sea carbonate dissolution [Zeebe et al., ‐ [11] The rationale for having the pre PETM fluxes ini- 2009]. The extended duration of the dissolution event could tially set higher than the modern values was to account for not be reproduced in the model without the continued carbon accelerated weathering influx during the Paleocene due to release. higher pCO2 and higher temperature. Carbonate rain to the [14] A pulsed carbon release versus a single input peak seafloor within the model is split into two broad locations: results in a slightly larger maximum Ca2+ change in our shelf/upper slope (<600 m) and lower slope/rise/plains study, 0.69% versus 0.41% respectively (Figure 2g; dotted ‐ ‐ ‐ (>600 m). The pre PETM ratio of shallow to deep car- line represents single input scenario). Thus, most of the bonate rain was increased relative to the modern because sea dissolution takes place during the initial phase and any level was higher, which promotes shallow water marine subsequent smaller input of carbon has an additional effect carbonate deposition (as has been documented for the on maximum seawater [Ca2+] change. For the case shown Paleocene and Early Eocene [e.g., Kiessling et al., 2003]). In here in which carbonate and silicate weathering feedbacks general, observations imply relatively less pelagic carbonate are weak, this subsequent, additional carbon input prolongs production during the early Cenozoic, which leads to a the interval of Ca2+ perturbation and causes a time lag shallower CCD compared to the modern in agreement with between the [Ca2+] peak (around 100 ky) and the initial the observations [Tyrrell and Zeebe, 2004]. Ravizza et al. carbon input peak. Increase in both carbonate and silicate – − [2001] estimate increased PETM weathering by 23% weathering rates are shown in Figure 2b (moles yr 1). Solid 26%, if the Os isotope excursion was solely due to higher lines (default nCC and nSi) and dashed lines (nCC and nSi are continental flux. The initial baseline atmospheric CO2 con- 0 0.9 and 0.7, respectively) compare the effect of carbonate centration is represented by pCO2 in the model and was set and silicate weathering parameters on the magnitude of at 1,000 ppm. During the course of the simulation, atmo- weathering. Figure 2c displays changes in total alkalinity spheric pCO2 is predicted by the model and as it varies, it (TA) and total carbon (TC). Simulated atmospheric pCO2 is affects the rate of weathering of carbonate and silicate rocks. shown in Figure 2f. All of the above mentioned parameters By default, silicate and carbonate weathering parameters 2− 2+ (TA, TC, [CO3 ], CCD, [Ca ]), which are summarized in (nSi and nCC) were set at 0.2 and 0.4, respectively. These Figure 2, pertain to the deep Pacific ocean. The deep Pacific values are at the lower end of the spectrum commonly used in ocean was arbitrarily chosen because it represents the largest the literature (conservative estimates) [Walker et al., 1981; basin. Moreover, the differences in [Ca2+] between the Berner and Kothavala, 2001; Zeebe et al., 2008; Zachos basins are negligible. 2+ et al., 2008; Uchikawa and Zeebe, 2008; Zeebe et al., 2009]. [15] The maximum deep ocean [Ca ] change (in percent) [12] We investigated the effects of increased weathering on 2+ under the three different conditions of increased weathering the [Ca ] inventory with three different sets of parameter (as described in section 1), as well as carbon input, are values. Scenario 1: enhanced carbonate weathering (nCC shown in Figures 3–5. Enhanced carbonate weathering varied between 0.4 and 1.4, nSi kept constant at 0.2), Scenario (Figure 3), enhanced silicate weathering (Figure 4) and 2: enhanced silicate weathering (nCC kept constant at 0.4, simultaneous increase of both carbonate and silicate nSi varied between 0.2 and 1.2), and Scenario 3: enhanced weathering (Figure 5), correspond to scenarios 1, 2 and 3, carbonate as well as silicate weathering (nCC and nSi simul- respectively. taneously varied between 0.4 and 1.4 and 0.2 and 1.2, [16] Under the weathering scenario 1 (n varied between respectively). The weathering parameterization used in this CC 0.4 and 1.4, nSi kept constant at 0.2), the simulated maxi- study is based on Uchikawa and Zeebe [2008]. mum deep ocean Ca2+ concentration increase (D[Ca2+]) ranges from 0.14% to 2.07% (Figure 3). D[Ca2+] is the ratio 4. Results of maximum [Ca2+] over the initial [Ca2+] expressed in D 2+ [13] The results presented in Figure 2 correspond to the percent. The magnitude of [Ca ] is most sensitive to the PETM model simulation in which carbonate and silicate amount of carbon released (ranges from 1,000 to 6,000 Pg), weathering parameters were kept constant at default values whereas the increased sensitivity of carbonate weathering has a minor effect on D[Ca2+]. Even when the carbonate (nCC = 0.4, nSi = 0.2) and the total carbon input was 3,000 Pg weathering parameter nCC is raised to 1.4 the change in

4of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 2. (a) PETM carbon release scenario. Here t = 0 corresponds to the onset of the PETM (Paleocene‐Eocene boundary). (b) Increase in both carbonate (green) and silicate weathering (magenta) rates. Solid lines represent default weathering feedback (nCC = 0.4 and nSi = 0.2). Dashed lines represent increased weathering feedback nCC = 0.9 and nSi = 0.7. This shows that the increased feedback results in 2− higher weathering rates. (c) TA (blue) and TC (red) of the Deep Pacific. (d) [CO3 ] of the Deep Pacific (black) and the Deep Atlantic (green). (e) Simulated CCD changes in the Pacific (black) and Atlantic 2+ (green) oceans. (f) Simulated atmospheric pCO2. (g) [Ca ] of the Deep Pacific. The solid line represents the change in Ca2+ as a result of initial plus continuous input shown in Figure 2a, whereas the dashed line represents the change in Ca2+ as result of a single input (s.i.). The green stars represent the points of max Ca2+ concentration.

5of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 3. Simulated change in [Ca2+] as a function of carbon input and increased carbonate weathering 12 −1 flux. The silicate weathering flux was kept constant at the initial value of 5.7 × 10 mol C yr ,nSi = 0.2.

[Ca2+] is insignificant (for constant carbon input). Similar to Figure 5 shows the magnitude of D[Ca2+] as a function of scenario 1, the weathering scenario 2 (nCC kept constant at carbon input (1,000 to 6,000 Pg) and enhanced carbonate and 2+ 0.4, nSi varied between 0.2 and 1.2) also shows that D[Ca ] silicate weathering rates (which result from increased nCC and strongly depends on the amount of carbon input. Under this nSi as follows from equations (5) and (6); see Figure 2b). As scenario, the increased silicate weathering also has a minor seen in the two other scenarios, the D[Ca2+] essentially only effect on D[Ca2+]. However, under this scenario, the D[Ca2+] depends on the amount of carbon input, with almost no visible decreases as the silicate weathering flux increases (Figure 4). effect from enhanced weathering. The D[Ca2+] ranges from The reason for this is that the increased silicate flux increases 0.07 to 1.51%. the total alkalinity (but it does not increase total carbon, unlike carbonate weathering) of the deep ocean, allowing the 5. Discussion CCD to deepen. This results in a larger area of the ocean supersaturated with respect to calcite preserving calcium [18] The model used in our study shows that dissolution carbonate falling from the surface (carbonate burial), which as well as enhanced carbonate and silicate weathering rates 2+ have a small effect on the ocean’s calcium inventory during consequently decreases the amount of dissolved Ca (see 2+ discussion for more information). the PETM, even though the rate of delivery of Ca from the land to the ocean probably increased significantly. The [17] Silicate and carbonate weathering both depend on reason for this is that after ∼10 ky enhanced weathering atmospheric CO2, thus it is unlikely that they increased independently from each other during the PETM. Therefore increases the calcite saturation state of the deep ocean. As a another set of simulations, which assumes concurrent consequence, the CCD deepens, leading to an increased increase in the sensitivity of carbonate and silicate weathering burial of calcite because a larger area of the seafloor is was conducted (weathering scenario 3). Under this scenario, bathed in supersaturated water [Zeebe and Westbroek, carbonate and silicate weathering sensitivity parameters were 2003]. Therefore, the process of increased weathering (adding Ca2+) is almost compensated for by the process of simultaneously varied over a range of values covering the 2+ most plausible scenarios (n was increased from 0.4 to 1.4 increased burial (removing Ca ) even though the com- CC pensation is not instantaneous (there is a time lag between and nSi was simultaneously increased between 0.2 and 1.2).

6of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 4. Simulated change in [Ca2+] as a function of carbon input and increased silicate weathering 12 −1 flux. The carbonate flux was kept constant at the initial value of 15.8 × 10 mol C yr ,nCC = 0.4. the response in the CCD to increased weathering rates). This Pacific ocean first slightly shallows and then deepens by time lag enables [Ca2+] to increase but this increase has an approximately 0.6 km (over the first 70 ky). Shoaling of the insignificant effect on the CCD change. Thus weathering CCD produces two effects, which both contribute to the has a minor effect on the change of calcium inventory over overall increase of [Ca2+]. First, as the CCD shoals, a bigger millennial timescales (Figure 5). fraction of the ocean floor is being exposed to undersaturated 2+ 2+ [19] On the other hand, the D[Ca ] strongly depends on water, which directly lessens the amount of Ca burial. the carbon input (Figure 5). As it was earlier explained and Because weathering continues, this introduces a source/sink shown in Figure 3, CO2 reacts in seawater to produce mostly imbalance. The second effect is also linked to the CCD. As the − HCO3. The input of CO2 (3,000 Pg C over the first 6 ky in CCD shallows and the saturation state of seawater with respect our model run) to the ocean increases the dissolved inor- to carbonate minerals starts to decline, dissolving carbonates 2− 2+ ganic carbon in seawater (Figure 2c). As a result, the [CO3 ] previously deposited to the sediments adds more Ca to the declines (Figure 2d), lowering the saturation state of sea- seawater. This process is also known as seafloor neutralization water with respect to carbonate minerals [Zeebe and Wolf‐ [Archer et al., 1997, 1998; Ridgwell and Hargreaves,2007]. Gladrow, 2001; Ridgwell and Zeebe, 2005]. This decrease The overall D[Ca2+] for the given scenario is illustrated in 2− in [CO3 ] can be illustrated by combining equations (3) Figure 2g. After about 60 ky when the carbon input ceases, the and (4) to get the following aqueous carbonate equilibrium calcium concentration still remains high (because of weak reaction: weathering feedback) and begins to return to the initial state after around 100 ky, facilitated by prolonged enhanced car- CO þ CO2 þ H O ! 2HCO ð7Þ 2ðÞaq 3 2 3 bonate and silicate weathering.

2− With increased CO2(aq), there is a removal of CO3 and an 5.1. Difference Between Carbonate 2− increase of bicarbonate. With a decrease of CO3 , the CCD and Silicate Weathering shoals. The global CCD (average CCD of all ocean basins) [20] The results presented in Figures 3 and 4 serve to shoals by approximately 0.75 km as a result of the carbon compare the separate impacts of increased carbonate weath- input (simulation shown in Figure 2), whereas the CCD in the ering (no change in FSi) and increased silicate weathering (no

7of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 5. Simulated change in [Ca2+] as a function of carbon input and simultaneous increase of both carbonate and silicate weathering fluxes.

2+ change in FC) on the D[Ca ], respectively. The simulation the net change of TC is zero, since every two moles of carbon shows that at constant carbon input with increasing carbonate absorbed on continents from the ocean‐atmosphere reservoir 2+ 2+ weathering, the D[Ca ] increases. On the contrary, the D[Ca ] in form of the CO2 return back to the ocean in the form of − decreases with increased FSi at constant carbon input. There two moles of HCO3 (note that in steady state, burial of are two processes that are responsible for the observed carbonate is balanced by volcanic input; Figure 6). However, incongruity. The first process is associated with the different ultimately silicate weathering of course sequesters carbon. 0 12 −1 magnitudes of the weathering fluxes FC (15.8 × 10 mol C yr ) The sequestration begins at the point when the net burial 0 12 −1 0 and FSi (5.7 × 10 mol C yr ). The initial FC is already more exceeds the net input of calcium ion and carbon is being 0 than twice FSi, thus the rate of delivery of calcium ions is sequestered in the form of CaCO3. One could also argue that larger with increased carbonate weathering as opposed to CO2 sequestration starts when net burial exceeds the baseline increased silicate weathering, producing a positive D[Ca2+]. riverine flux. In our model these two events occur approxi- The second and more important process that results in a mately at the same time, around 100 ky in the standard run. 2+ different trend of D[Ca ] as a function of carbonate and [21] With increased silicate weathering, the TC/TA ratio silicate weathering is due to different influences of the sili- changes, increasing the alkalinity of seawater. This means cate and carbonate weathering on total carbon in the ocean. ‐ One mole of carbonate delivered from the land to the ocean Table 1. Net Change (Mole) in Flux per One Mole of Weathered atmosphere system adds one mole of calcium (on a net Ca2+ for Combined Ocean‐Atmosphere Systema basis), one mole of TC and two moles of TA (equation (1) and Table 1). On the other hand, on this timescale silicate Carbonate Silicate weathering does not affect the TC but has the same effect as Ca2+ +1 +1 carbonate weathering on TA and calcium (equation (2) and TC +1 0 Table 1). The reason why TC is not affected by silicate TA +2 +2 a weathering is that the atmosphere and ocean act as a single See Figure 6 and equations (1) and (2). Only FC − Fa and FSi − Fa, reservoir on timescales of several thousands of years. Thus, respectively, are considered. TC is total carbon, and TA is total alkalinity.

8of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Figure 6. Carbonate and silicate weathering fluxes. Compare with Table 2. X is total inorganic carbon, and Y is total alkalinity. On long timescales, the atmospheric and oceanic box act as a single reservoir, which is indicated by the dashed lines.

9of13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

2+ a Table 2. The Effect of FC and FSi on the CCD and D[Ca ] 0 −1 0 −1 2+ Experiment FC (mol C yr )FC FSi (mol C yr )FSi D[Ca ](%) DCCD (km) 1 15.8 21.5 5.7 5.7 0.10 0.46 2 15.8 15.8 5.7 11.4 0.04 0.75

a 12 −1 After increasing both weathering rates independently by 5.7 × 10 mol C yr (Experiment 1: FC increased; Experiment 2: FSi increased), the effect of carbonate weathering increase on the CCD change is much less than that of silicate weathering. that silicate weathering has a larger per‐mole effect on the the PETM was probably in the range of 0.07% to 0.7% deepening of the CCD than carbonate weathering. This (Figure 5, area below 3,000 Pg C). effect is illustrated in Table 2. As mentioned above, silicate [24] Even if the most extreme scenarios is assumed, when and carbonate weathering were most likely simultaneously D[Ca2+] is around 2% (Figure 3, top right corner) and when enhanced during the PETM (scenario 3). The effect of such this change is applied to the equation for Mg/Ca ratio to an increase on D[Ca2+] is shown in Figure 5 and essentially temperature conversion of Anand et al. [2003], the calcu- represents the sum of the scenarios shown in Figures 3 lated temperature uncertainty is approximately 0.2°C, which and 4. Among the three conditions, this is the one in which is within the uncertainty of ±1.2 to ±1.4°C given by Anand D[Ca2+] is the least dependent on the amount of increased et al. [2003] and Dekens et al. [2002]. If the same change in 2− weathering. calcium is applied to the equation for calculating the D[CO3 ] from B/Ca ratios of Yu and Elderfield [2007] the uncertainty 5.2. Effect of Ca Inventory Changes −1 is ∼4 mmol kg (C. wuellerstorfi species), which is within on B/Ca and Mg/Ca −1 their uncertainty envelope of ±10 mmol kg . This leads to the [22] Deep ocean carbonate saturation and temperature of conclusion that the PETM carbon cycle perturbation was not past climate events have been reconstructed using the ratios large enough to disturb the calcium ion concentration to the of boron to calcium and magnesium to calcium, respectively extent that would question the feasibility of using B/Ca for [Yu and Elderfield, 2007; Elderfield and Ganssen, 2000]. reconstructing past variations in deep water carbonate satu- One assumption used in those methods is that the concen- ration state or to question the reliability of the Mg/Ca method 2+ tration of Ca does not fluctuate on timescales shorter than for reconstructing paleotemperatures. Nevertheless, there are the residence time of calcium. Here we investigate whether other processes that could affect the reliability of the above the magnitude of carbon perturbation that occurred during the mentioned methods (such as dissolution on the seafloor, 2+ PETM was capable of changing the Ca concentration diagenesis, vital effects, etc.), but they are beyond the scope enough to question the reliability of the above mentioned of this study. methods. [23] The summary of our simulations is illustrated in 5.3. Effect on the Ocean Calcium Isotope Budget Figure 5. The results represented here are derived from a [25] It has been speculated that throughout Earth’shistory, variety of different scenarios in which the magnitude of the isotopic composition of the marine calcium has varied silicate and carbonate weathering fluxes were simulta- considerably. This change is mainly driven by the imbalance neously increased. The scale of carbon input ranged from between riverine and hydrothermal inputs of Ca2+ as well as 1,000 to 6,000 Pg C. This approach allows us to include the outputs via sedimentation of CaCO3 on the seafloor [e.g., plausible Ca changes (D[Ca2+]) during the PETM, thus giv- De La Rocha and DePaolo, 2000; Schmitt et al., 2003; Heuser ing us a range of scenarios. The D[Ca2+] is greatest under the et al., 2005]. As we showed here, such an inequality might scenario where the input of carbon is 6,000 Pg C, regardless have altered the seawater [Ca2+]byupto2%duringthePETM of the weathering sensitivity used (Figure 5). Under this (under the most extreme scenario) and consequently could scenario, Ca2+ increases by approximately 1.5%. Further- cause an excursion of the seawater Ca isotope composition 44 44 more, according to Zeebe et al. [2009], the initial PETM (d Casw). Another factor responsible for changes in d Casw is carbon input was probably 3,000 Pg C or less. Also, Zeebe the isotopic composition of the minerals that are being 44 et al. [2009] calculated an increase in atmospheric CO2 weathered (d Caw) as silicate and carbonate rocks have dif- from a baseline of 1,000 ppmv to approximately 1,700 ppmv ferent d44Ca values. Accordingly, if the input of calcium from during the onset of the PETM. The increase in CO2 can be these weathering processes changed in the past, it would result 44 compared to the results of the study by Uchikawa and Zeebe in a different d Caw. In their study, Griffith et al. [2008] used 44 [2008], which covered the possible range of weathering values of d Caw ranging from −1.8l to −1.2‰.Herewecover 44 fluxes as a function of pCO2 as given by Berner and Kothavala a wider range of possible d Caw scenarios (−2.0‰ to −1.0‰). [2001] and Walker and Kasting [1992]. Applying these find- Our simulation is based on the following mass balance of the ings to the estimated increase of pCO2 during the main stage Ca isotopic composition of seawater: of the PETM (from 1,000 ppmv to 1,700 ppmv) suggests that 44 ÀÁÀÁ the silicate weathering flux did not increase by more than 50% dðÞ Casw 44 44 44 NCa ¼ Fin Caw Casw F out D Caout ð8Þ regardless of the model employed. The increase in carbonate dt weathering caused by the 700 ppm rise in pCO2 is not more 2+ where NCa is the amount of Ca in the ocean in moles, Fin is than 100% in any scenario presented by Uchikawa and Zeebe 2+ [2008]. By using the above mentioned results (carbon input the flux of Ca to the ocean from both silicate and carbonate ≤ weathering and hydrothermal inputs, Fout represents the flux 3,000 Pg C and weathering increase <100%) and comparing 2+ D 2+ of Ca from the ocean due to carbonate sedimentation in mol it to the results of our study, [Ca ] during the main stage of −1 44 yr and D Caout is the difference between the Ca isotopic

10 of 13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

44 44 2+ Figure 7. Simulated change in d Casw as a function of d Caw and increase in [Ca ].

composition of marine carbonates and seawater in ‰.In suggests significant imbalances between input and output of 2+ steady state, Fin equals Fout and these values are initially set to Ca from the ocean over the Cenozoic, which would be one 12 −1 0 0 21.5 × 10 mol yr (FC + FSi). In our study, we kept possible explanation for variable calcium isotopic compo- 44 D Caout constant at −1.3‰ [De La Rocha and DePaolo, sition over this time period [De La Rocha and DePaolo, 2000; Fantle and DePaolo, 2005; Griffith et al., 2008]. We 2000; Fantle, 2010]. However, these processes, such as 44 solved equation (8) numerically for d Casw, varying Fin in dolomite formation and decoupling of calcium riverine flux order to produce increases in NCa and thus the corresponding from the carbonate flux, affect seawater concentrations on 2+ 44 change in [Ca ] (Figure 7). The largest change in d Casw multimillion year timescales. 44 2+ (Dd Casw) is produced when the D[Ca ] is 2% and when the Ca isotopic composition of the weathered material is 6. Conclusions 44 −2‰. Under this scenario, the Dd Casw is around −0.06‰. As previously mentioned, because the D[Ca2+] during the [27] In summary, our results show that dissolution and PETM was most likely less than 0.6%, it follows that the enhanced continental weathering had a small effect on the ’ 2+ change in isotopic composition of seawater during the PETM ocean sCa inventory during the PETM. Additional simu- − ‰ lations showed that this conclusion also holds at modern was most likely less than 0.04 . These results show that the 2+ 2+ Ca isotopic composition of seawater was essentially not seawater Ca and Mg concentrations. The small effect that affected by the carbon perturbation, it remained nearly con- is seen in this study differs among the carbonate and silicate stant over the PETM. weathering fluxes. Accelerated carbonate fluxes increase the D[Ca2+] of the ocean. On the contrary, enhanced silicate [26] Variations in the marine calcium cycle throughout the 2+ Earth’s history have been reported but the factors respon- fluxes show a tendency to decrease the [Ca ] on the given sible for the variations are only important over longer timescale because increased silicate flux increases the alkalinity of seawater (TC/TA ratio changes), resulting in a larger timescales (several hundreds of thousands to millions of ‐ years). Yet it is improbable that long‐term processes per mole effect on the deepening of the CCD than carbonate affected the PETM (timescale ∼200 ky). One hypothesis weathering. Since the effect of the silicate weathering over-

11 of 13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211 whelms the effect of the carbonate weathering, the overall The calculated value of 0.20% for D[Ca2+] is a lower estimate result is that with the simultaneous increase of both fluxes, the since it does not account for the variable porosity and the D[Ca2+] decreases or remains nearly constant. This is erosion of the sediments. Since the porosity and the erosion of reflected in the almost horizontal lines in Figure 5. On the the sediments in our model vary, the ratio of total CaCO3 other hand, the D[Ca2+] is highly dependent on the size of the available during erosion to the mass contained in the original 2+ carbon input. This implies that the effect of delivery of Ca layer (MCaCO3) is given by: from the dissolution on the seafloor is much greater on the 1 f 0 ocean Ca inventory than the effect that results from the 1 þ 0 c ðA3Þ 0 continental runoff. By using the results from this study and 1 1 1 fc combining them with the findings from two other studies [Uchikawa and Zeebe, 2008; Zeebe et al., 2009], we were where 0 and 1 are the porosities of a pure clay and calcite 2+ − − able to constrain the most plausible range of D[Ca ]. The layer, respectively. The fraction (1 0)/(1 1) is 0.4 in our 2+ 0 maximum increase of Ca was most likely less than 0.7% calculation [Zeebe and Zachos, 2007]. Because f c was relative to the concentration prior to the PETM; the maximum assumed to be 0.71, the fraction in the above equation is equal 2+ D[Ca ] is observed around 100 ky after the onset of the to 4. From equation (A3) it follows that the CaCO3 available event when the carbon input ceases. Thus, the calculated during erosion is approximately 2 times greater than the mass temperature uncertainty (for the most extreme scenario, in the original surface layer (equation (A1)). Thus, the cor- 2+ 2− D 2+ D[Ca ] = 2%) of 0.2°C and D[CO3 ] uncertainty of ∼4 mmol rected [Ca ] is actually 2 times larger and is approximately −1 44 ’ kg become even smaller; the Dd Casw is in the range from equal to 0.40% of the ocean s Ca inventory. +0.01‰ to −0.04‰. [30] Acknowledgments. We are grateful to A. Ridgwell and two [28] We conclude that the magnitude of calcium ion anonymous reviewers for their constructive and thorough reviews, which change due to enhanced weathering and carbonate dissolu- improved the manuscript. We also want to thank the handling editor, tion during the PETM is not large enough to affect the C. Charles. This research was supported by NSF grant OCE09-02869 to accuracy of proxies based on B/Ca and Mg/Ca ratios and R.E.Z. 44 2+ d Casw via Ca inventory changes. On multimillion timescales, however, other processes control seawater calcium References fluctuations such as dolomite formation [Holland and Anand, P., H. Elderfield, and M. H. Conte (2003), Calibration of Mg/Ca thermometry in planktonic foraminifera from a sediment trap time series, Zimmermann, 2000] and decoupling of calcium riverine Paleoceanography, 18(2), 1050, doi:10.1029/2002PA000846. flux from the carbonate flux [Wallmann, 2001; Heuser et al., Archer, D., H. Kheshgi, and E. Maier‐Reimer (1997), Multiple timescales – 2005]. for neutralization of fossil fuel CO2, Geophys. Res. Lett., 24, 405 408. Archer, D., H. Kheshgi, and E. Maier‐Reimer (1998), Dynamics of fossil fuel CO2 neutralization by marine CaCO3, Global Biogeochem. Cycles, Appendix A: Estimating the Effect of Dissolution 12, 259–276. on Seawater Calcium Berner, R. A., and Z. Kothavala (2001), GEOCARB III: A revised model of atmospheric CO2 over Phanerozoic time, Am. J. Sci., 304, 397–437. ‐ [29] In order to estimate the volume of the sediment Berner, R. A., A. C. Lasaga, and R. M. Garrels (1983), The carbonate silicate geochemical cycle and its effect on atmospheric carbon dioxide available for dissolution it is necessary to make a few over the past 100 million years, Am. J. Sci., 283, 641–683. assumptions. First, we assumed that the area of the entire Bice, K. L., and J. Marotzke (2002), Could changing ocean circulation have ocean floor is equal to the area of the ocean surface (Aoc), destabilized methane hydrate at the Paleocene/Eocene boundary?, Paleo- which is 3.49 × 1014 m2. If the CCD is at 3.5 km, then the area ceanography, 17(2), 1018, doi:10.1029/2001PA000678. Bowen, G. J., D. J. Beerling, P. L. Koch, J. C. Zachos, and T. Quattlebaum of the ocean floor above the CCD is approximately 30% of (2004), A humid climate state during the Paleocene‐Eocene thermal maxi- the entire ocean floor. A top sediment mixed layer of 10 cm mum, Nature, 432,495–499. −3 Dekens, P. S., D. W. Lea, D. K. Pak, and H. J. Spero (2002), Core top cal- was assumed (hs). The density of sediments r = 2.5 g cm . ibration of Mg/Ca in tropical foraminifera: Refining paleotemperature fc is the carbonate fraction in dry sediments and a value of estimation, Geochem. Geophys. Geosyst., 3(4), 1022, doi:10.1029/ 0.71 was chosen (calculated by LOSCAR). The porosity of 2001GC000200. the sediments () is 0.7. In order to calculate the mass De La Rocha, C. L., and D. J. DePaolo (2000), Isotopic evidence for variations in the marine calcium cycle over the Cenozoic, Science, 289, of CaCO3 in sediments above the CCD (MCaCO ) the 3 1176–1178. following equation was employed: Dickens, G. R. (2000), Methane oxidation during the late Palaeocene thermal maximum, Bull. Soc. Geol. Fr., 171(1), 37–49. M ¼ Aoc 0:3 h ðÞ1 f ðA1Þ CaCO3 s c Dickens, G. R. (2001), Carbon addition and removal during the Late Palaeocene thermal maximum: Basic theory with a preliminary treatment 18 of the isotope record at ODP Site 1051, Blake Nose, in Western North This gives an amount of 5.6 × 10 g of CaCO3 which is equal 16 Atlantic Palaeogene and Cretaceous Palaeoceanography, edited by to 5.6 × 10 moles of C (∼670 Pg C) and as many moles of D. Kroon, R. D. Norris, and A. Klaus, Geol. Soc. Spec. Publ., 183, 2+ – Ca (mCa) since the molar ratio of carbon to calcium is 1:1. In 293 305. order to get the percent change of [Ca2+] that such an input Dickens, G. R., J. R. O’Neil, D. K. Rea, and R. M. Owen (1995), Dissoci- ation of oceanic methane hydrate as a cause of the carbon isotope excur- would produce, two more pieces of information are neces- – 21 sion at the end of the Paleocene, Paleoceanography, 10, 965 971. sary: the mass of the ocean (Moc = 1.38 × 10 kg) and the Dunkley Jones, T., A. Ridgwell, D. J. Lunt, M. A. Maslin, D. N. Schmidt, seawater calcium concentration at the time of dissolution, and P. J. Valdes (2010), A Paleogene perspective on climate sensitivity and methane hydrate instability, Philos. Trans. R. Soc. A, 368, 2395–2415, which was taken as 20 mmol for the PETM ([Ca]sw)[Tyrrell 2010. and Zeebe, 2004]: Elderfield, H., and G. Ganssen (2000), Past temperature and d18O of sur- ÂÃ m face ocean waters inferred from foraminiferal Mg/Ca ratios, Nature, D Ca2þ ¼ Ca 100 ¼ 0:20% ðA2Þ 405, 442–445. Moc ½Ca sw Fantle, M. S. (2010), Evaluating the Ca isotope proxy, Am. J. Sci., 310, 194–230.

12 of 13 PA3211 KOMAR AND ZEEBE: OCEAN CALCIUM ION CHANGE DURING THE PETM PA3211

Fantle, M. S., and D. J. DePaolo (2005), Variations in the marine Ca cycle Thomas, D. J., T. J. Bralower, and C. E. Jones (2003), Neodymium isotopic over the past 20 million years, Earth Planet. Sci. Lett., 237, 102–117. reconstruction of late Paleocene‐early Eocene thermohaline circulation, Griffith, E. M., A. Paytan, K. Caldeira, T. D. Bullen, and E. Thomas Earth Planet. Sci. Lett., 209, 309–322. (2008), A dynamic marine calcium cycle during the past 28 million Thomas, E. (2007), Cenozoic mass extinctions in the deep sea; what dis- years, Science, 322, 1671–1674. turbs the largest habitat on Earth?, in Large Ecosystem Perturbations: Heuser, A., A. Eisenhauer, F. Böhm, K. Wallmann, N. Gussone, P. N. Causes and Consequences, edited by S. Monechi, R. Coccioni, and Pearson, T. F. Nägler, and W.‐C. Dullo (2005), Calcium isotope (d44/40Ca) M. Rampino, Spec. Pap. Geol. Soc. Am., 424,1–24. variations of Neogene planktonic foraminifera, Paleoceanography, 20, Tyrrell, T., and R. E. Zeebe (2004), History of carbonate ion concentration PA2013, doi:10.1029/2004PA001048. over the last 100 million years, Geochim. Cosmochim. Acta, 68(17), Holland, H. D., and H. Zimmermann (2000), The dolomite problem revis- 3521–3530. ited, Int. Geol. Rev., 42, 481–490. Uchikawa, J., and R. E. Zeebe (2008), Influence of terrestial weathering on Horita, J., H. Zimmermann, and H. D. Holland (2002), Chemical evolution ocean acidification and the next glacial inception, Geophys. Res. Lett., of seawater during the Phanerozoic: Implications from the record of 35, L23608, doi:10.1029/2008GL035963. marine evaporites, Geochim. Cosmochim. Acta, 66(21), 3733–3756. Walker, J. C. G., and J. F. Kasting (1992), Effects of fuel and forest con- Kelly, D. C., J. C. Zachos, T. J. Bralower, and S. A. Schellenberg (2005), servation on future levels of atmospheric carbon dioxide, Palaeogeogr. Enhanced terrestrial weathering/runoff and surface ocean carbonate pro- Palaeoclimatol. Palaeoecol., 97, 151–189. duction during the recovery stages of the Paleocene‐Eocene thermal max- Walker, J. C. G., P. B. Hays, and J. F. Kasting (1981), Negative feedback imum, Paleoceanography, 20, PA4023, doi:10.1029/2005PA001163. mechanism for the long‐term stabilization of Earth’s surface temperature, Kennett, J. P., and L. D. Stott (1991), Abrupt deep‐sea warming, palaeocea- J. Geophys. Res., 86, 9776–9782. nographic changes and benthic extinctions at the end of the Palaeocene, Wallmann, K. (2001), Controls of hte Cretaceous and Cenozoic evolution of Nature, 353, 225–229. seawater composition, atmospheric CO2 and climate, Geochim. Cosmochim. Kiessling, W., E. Flügel, and J. Golonka (2003), Patterns of phanerozoic Acta, 65, 3005–3025. carbonate platform sedimentation, Lethaia, 36(3), 195–225. Webb, A., L. R. Leighton, S. A. Schellenberg, E. A. Landau, and E. Thomas Koch, P. L., J. C. Zachos, and P. D. Gingerich (1992), Correlation betwen (2009), Impact of the Paleocene‐Eocene thermal maximum on deep‐ocean isotope records in marine and continental carbon reservoirs near the microbenthic community structure: Using rank‐abundance curves to quan- Paleocene/Eocene boundary, Nature, 358, 319–322. tify paleoecological response, Geology, 37, 783–786. Lea, D. W., D. K. Pak, and H. J. Spero (2000), Climate impact of late Quater- White, A. F., and A. F. Blum (1995), Effects of climate on chemical weath- nary equatorial Pacific sea surface temperature variations, Science, 289, ering in watersheds, Geochim. Cosmochim. Acta, 59, 1729–1747. 1719–1724. Yu, J., and H. Elderfield (2007), Benthic foraminiferal B/Ca ratios reflect Lowenstein, T. K., M. N. Timofeeff, S. T. Brennan, L. A. Hardie, and R. V. deep water carbonate saturation state, Earth Planet. Sci. Lett., 258,73–86. Demicco (2001), Oscillation in Phanerozoic seawater chemistry: Evidence Yu, J., H. Elderfield, and B. Hönisch (2007), B/Ca in planktonic foraminif- from fluid inclusions, Science, 294, 1086–1088. era as a proxy for surface seawater pH, Paleoceanography, 22, PA2202, Morse, J. W., and F. T. Mackenzie (1990), Geochemistry of Sedimentary doi:10.1029/2006PA001347. Carbonates, Dev. Sedimentol., vol. 48, Elsevier, Amsterdam. Zachos, J. C., M. Pagani, L. Sloan, E. Thomas, and K. Billups (2001), Pagani, M., K. Calderia, D. Archer, and J. C. Zachos (2006), An ancient Trends, rhythms, and aberrations in global climate 65 Ma to present, carbon mystery, Science, 314, 1556–1557. Science, 292, 686–693. Panchuk, K., A. Ridgwell, and L. R. Kump (2008), Sedimentary response Zachos, J. C., M. W. Wara, S. Bohaty, M. L. Delaney, M. R. Petrizzo, to Paleocene‐Eocene thermal maximum carbon release: A model‐data A. Brill, T. J. Bralower, and I. Premoli‐Silva (2003), A transient rise in comparison, Geology, 36(4), 315–318, doi:10.1130/G24474A.1. tropical sea surface temperature during the Paleocene‐Eocene thermal Ravizza, G., R. N. Norris, J. Blusztajn, and M.‐P. Aubry (2001), An maximum, Science, 302, 1551–1554. osmium isotope excursion associated with the late Paleocene thermal max- Zachos, J. C., et al. (2005), Rapid acidification of the ocean during the imum: Evidence of intensified chemical weathering, Paleoceanography, Paleocene‐Eocene thermal maximum, Science, 308, 1611–1615. 16, 155–163. Zachos, J. C., G. R. Dickens, and R. E. Zeebe (2008), An early Cenozoic Ridgwell, A., and J. C. Hargreaves (2007), Regulation of atmospheric CO2 perspective on greenhouse warming and carbon‐cycle dynamics, Nature, by deep‐sea sediments in an Earth system model, Global Biogeochem. 451, 279–283, doi:10.1038/nature06588. Cycles, 21, GB2008, doi:10.1029/2006GB002764. Zeebe, R. E., and P. Westbroek (2003), A simple model for the CaCO3 satu- Ridgwell, A., and D. N. Schmidt (2010), Past constraints on the vulnerabil- ration state of the ocean: The “Strangelove,” the “Neritan,” and the “Cretan” ity of marine calcifiers to massive carbon dioxide release, Nat. Geosci., 3, Ocean, Geochem. Geophys. Geosyst., 4(12), 1104, doi:10.1029/ 196–200. 2003GC000538. Ridgwell, A., and R. E. Zeebe (2005), The role of the global carbonate Zeebe, R. E., and D. A. Wolf‐Gladrow (2001), CO2 in Seawater: Equilibrium, cycle in the regulation and evolution of the Earth System, Earth Planet. Kinetics, Isotopes, Elsevier Oceanogr. Ser., vol. 65, 346 pp., Elsevier, Sci. Lett., 234, 299–315. Amsterdam. Robert, C., and J. P. Kennett (1994), Paleocene and Eocene kaolinite dis- Zeebe, R. E., and J. C. Zachos (2007), Reversed deep‐sea carbonate ion tribution in the South Atlantic and Southern Ocean: Antarctic climatic basin gradient during Paleocene‐Eocene thermal maximum, Paleoceano- and paleoceanographic implications, Mar. Geol., 103,99–110. graphy, 22, PA3201, doi:10.1029/2006PA001395. Rosenthal, Y., E. A. Boyle, and N. Slowey (1997), Environmental controls Zeebe, R. E., J. C. Zachos, K. Caldeira, and T. Tyrrell (2008), Oceans: Carbon on the incorporation of Mg, Sr, F and Cd into benthic foraminifera shells emissions and acidification, Science, 321,51–52, doi:10.1126/science. from Little Bahama Bank: Prospects for thermocline paleoceanography, 1159124. Geochim. Cosmochim. Acta, 61, 3663–3643. Zeebe, R. E., J. C. Zachos, and G. R. Dickens (2009), Carbon dioxide forcing Schmitt, A. D., F. Chabaux, and P. Stille (2003), The calcium riverine and alone insufficient to explain Palaeocene‐Eocene thermal maximum warm- hydrothermal isotopic fluxes and the oceanic calcium mass balance, ing, Nat. Geosci., 2, 576–580, doi:10.1038/NGEO578. Earth Planet. Sci. Lett., 213, 503–518. Schmitz, B., and V. Pujalte (2007), Abrupt increase in seasonal extreme N. Komar and R. E. Zeebe, Department of Oceanography, School of precipitation at the Paleocene‐Eocene boundary, Geology, 35, 215–218. Ocean and Earth Science and Technology, University of Hawaii at Sluijs, A., et al. (2006), Subtropical Arctic Ocean temperatures during the – Manoa, 1000 Pope Rd., MSB 507, Honolulu, HI 96822, USA. (komar@ Palaeocene/Eocene thermal maximum, Nature, 441, 610 613. hawaii.edu)

13 of 13